To develop and validate an interpretable prediction model for delayed diagnosis of benign paroxysmal positional vertigo (BPPV).
Audiologists and ENT clinicians should be aware that delayed BPPV diagnosis remains a documented clinical problem; while this specific model is not yet validated for external clinical use, the identified risk factors may prompt earlier targeted assessment in high-risk patients.
Delayed diagnosis of BPPV leads to unnecessary patient suffering and healthcare costs; an interpretable predictive model could help triage at-risk patients earlier in primary care settings.
- 01Retrospective machine-learning model built and temporally validated on BPPV patients from Beijing.
- 02Model is described as 'interpretable,' supporting practical clinical adoption if externally validated.
- 03Delayed BPPV diagnosis is the primary outcome being predicted.
- 04Retrospective single-centre design limits immediate generalizability outside the study population.
- 05Published in Preventive Medicine Reports; vestibular/audiology relevance is direct.
An interpretable machine-learning model can predict delayed diagnosis of BPPV using routinely available clinical data.
studypartially supportedDelayed diagnosis of BPPV is a measurable and predictable clinical outcome.
studysupported- PMID
- 42369378
- DOI
- 10.1016/j.pmedr.2026.103543.
- Journal
- Preventive Medicine Reports
- Publication type
- research_article
- Evidence level
- 4
- Population
- Patients diagnosed with benign paroxysmal positional vertigo (BPPV) at hospitals in Beijing, China
- Intervention
- Interpretable machine-learning prediction model for delayed BPPV diagnosis
- Comparator
- Temporal validation cohort (split by time period)
Primary outcomes
Prediction of delayed diagnosis of BPPV; Model discrimination and calibration in temporal validation set